Motivation: The motivation for the current work arises from the limitations of 3D multi-slab diffusion-weighted imaging (3D-msDWI), particularly its inability to support advanced microstructure studies due to its long volume acquisition times (VAT). Goal(s): Develop a novel high SNR-efficiency signal generation and reconstruction framework based on 3D-msDWI that supports advanced microstructure studies at high spatial resolution. Approach: Substantial reduction in VAT is achieved by sub-sampling 3D k-space. An aliasing-free reconstruction is designed by incorporating deep-learned priors to recover the volume data. Results: The resulting framework is shown to generate 1mm isotropic 3-shell 66-direction data in 13min for the first time using 3D-msDWI acquisitions. Impact: Addressing the long VAT of 3D-msDWI can provide a new technique to generate high SNR data that supports advanced microstructure studies.
Lee et al. (Tue,) studied this question.